fche changed the topic of #systemtap to: http://sourceware.org/systemtap; email systemtap@sourceware.org if answers here not timely, conversations may be logged
orivej has quit [Ping timeout: 240 seconds]
<kerneltoast> gonna run the testsuite on that bad boy. it passes hrtimers_onthefly.exp atm
<kerneltoast> so it looks promising
<fche> for the onthefly tests, I recommend running them all
<fche> i.e.,
<fche> RUNTESTFLAGS="pr14546.exp FOO_onthefly.exp BAR_onthefly.exp BAZ_onthefly.exp"
<fche> because the problems seem to manifest themselves more reliably when these things are operated in sequence
<kerneltoast> want me to run those in a loop?
<fche> ideally
<irker170> systemtap: fche systemtap.git:master * release-4.4-23-ge8c7a2067 / testsuite/systemtap.base/pr14546.stp: testsuite pr14536.stp: toughen
<fche> strong typo game in that one
<irker170> systemtap: fche systemtap.git:master * release-4.4-24-g3e0d449fe / : configury noise: regenerate using fedora-33 auto*
<fche> kerneltoast, btw I like the bulkmode patch
<fche> lots of red in the diff, good
<fche> but
<fche> I bet you don't need that pthread_mutex* around the write(2) over on line 516 of the patch
<kerneltoast> how much u wanna bet
<fche> if they're using the same fd, the kernel will not interleave writes, I believe
<fche> how much
<fche> hmmmmmm
<fche> THIS MUCH!
<kerneltoast> tree fiddy?
<fche> freedom costs a buck oh five
<fche> btw there are static initializers for userspace pthread mutexes, so you don't really need that explicit funcall in init_relayfs()
<fche> anywho I must play ping-pong with the wife (seriously)
<fche> but yeah the patch looks good.
<kerneltoast> been a while since i've written userspace code
<kerneltoast> it was too easy
<kerneltoast> i gotta go back into my kernel hole
<kerneltoast> oh cool, write is atomic
<kerneltoast> what do i know about posix
<kerneltoast> fche, i'm running RUNTESTFLAGS="pr14546.exp FOO_onthefly.exp BAR_onthefly.exp BAZ_onthefly.exp" in a nonstop loop
<fche> yeah just not FOO and BAR literally mon ami
<fche> write is atomic
<fche> OF COURSE IT IS
<fche> trust the frank
<fche> affection.exp hrtimer_onthefly.exp kprobes_onthefly.exp
<fche> onthefly_common.tcl tracepoint_onthefly.exp uprobes_onthefly.exp
<kerneltoast> i'm too tired, gimme real commands
<kerneltoast> i'll just run the whole testsuite actually
<fche> sudo make installcheck RUNTESTFLAGS="pr14546.exp affection.exp hrtimer_onthefly.exp kprobes_onthefly.exp onthefly_common.tcl tracepoint_onthefly.exp uprobes_onthefly.exp"
<fche> go clean your room
<fche> to thine own self be true
<kerneltoast> you lost against your wife in ping-pong didn't you
<kerneltoast> and you're taking out your RAGE on innocent me
<fche> you are not wrong
<fche> she is cruel and vicious a competitor
<fche> looks pretty good. suggest simpler summary line, something like 'always use per-cpu bulkmode relayfs files to communicate with userspace' ?
<kerneltoast> good?
<fche> it looks good
<fche> if it survives a couple of rounds of that testing, ship it
<fche> may it be an early christmas
<kerneltoast> don't wanna just shippit and see what remains of the buildbots?
<fche> the buildbots are still a bit sore about the week long red here and there
<fche> I want to coddle them and comfort them this time
<kerneltoast> they can take it
<kerneltoast> fche, your command ain't working
<kerneltoast> gimme better commands
<kerneltoast> /bin/sh: systemtap.sum: No such file or directory
<fche> hm, drop the .tcl one I think
<fche> tell ya what, push your code to a branch, I'll also run some tests here
<kerneltoast> okey
<kerneltoast> gimme a branch name
<fche> branch/name
<kerneltoast> k
<irker170> systemtap: sultan systemtap.git:sultan/bulkmode * release-4.4-23-g024d1fad3 / runtime/print_flush.c runtime/transport/relay_v2.c runtime/transport/transport.c runtime/transport/transport.h staprun/relay.c: always use per-cpu bulkmode relayfs files to communicate with userspace
<irker170> systemtap: sultan systemtap.git:sultan/bulkmode * release-4.4-24-gfcd798b46 / tapset-timers.cxx: Revert "REVERTME: tapset-timers: work around on-the-fly deadlocks caused by mutex_trylock"
<kerneltoast> fche, thar's yer branch
<fche> ok
<fche> btw that stp_print_*lock business needs to be duplicated lightly to the runtime/dyninst/* directory, else stap --runtime=dyninst won't build
<fche> betcha kernel people such as yourself don't have the magic of dyninst-devel installed, thus your stap lacks this neat facility
<fche> I'll fix that myself
<irker170> systemtap: fche systemtap.git:sultan/bulkmode * release-4.4-25-gfd12d80b6 / runtime/dyninst/print.c runtime/dyninst/runtime_defines.h: dyninst transport: add _stp_print_*lock_irq* stubs
* kerneltoast doesn't know what dyninst even is
<fche> who even are you
<fche> I mean come on
<kerneltoast> i also don't know how to write a stap script
<fche> pure userspace binary-rewriting
<kerneltoast> but that's a secret
<fche> likewise
<kerneltoast> glad it's not just me
<fche> the last new systemtap script was written in 2006
<fche> all the rest are neural network based chatterbot outputs
<fche> tell no one
derek088_ has joined #systemtap
kerneltoast is now known as no_one
<no_one> why hello
<fche> whoa
<no_one> i just heard from this kerneltoast fellow
<no_one> about neurotic networks writing stap scripts
derek0883 has quit [Ping timeout: 260 seconds]
no_one is now known as kerneltoast
<kerneltoast> did i miss anything?
<fche> no one said anything
<kerneltoast> anything is quite vague
derek088_ has quit [Ping timeout: 264 seconds]
<fche> is it really
<kerneltoast> just like how dimensionality is a function of consciousness
derek0883 has joined #systemtap
<fche> wonder if I get thinner while I sleep <<< look what you made me do
<kerneltoast> depends on how long you sleep
<kerneltoast> abraham lincoln is quite thin these days
<kerneltoast> it's a new diet
<fche> stop or you'll make me cue Decomposing Composers from python
<kerneltoast> what about
<kerneltoast> Decomposing Composers Composing Compost
derek0883 has quit [Ping timeout: 264 seconds]
derek0883 has joined #systemtap
mjw has quit [Ping timeout: 246 seconds]
mjw has joined #systemtap
derek0883 has quit [Ping timeout: 260 seconds]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
khaled has quit [Quit: Konversation terminated!]
<fche> kerneltoast, uprobes_onthefly.exp still struck dead by the code
<fche> pity
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
irker170 has quit [Quit: transmission timeout]
<kerneltoast> fche, oof
<kerneltoast> well it's not the mutex_trylock at least
<kerneltoast> we really gotta get some vmcores from the buildbots
<kerneltoast> I'm running out of bugfixes to punt at you in the hopes of it fixing the buildbots
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
orivej has joined #systemtap
_whitelogger has joined #systemtap
mjw has quit [Quit: Leaving]
beauty2 has joined #systemtap
beauty1 has quit [Ping timeout: 272 seconds]
khaled has joined #systemtap
orivej has quit [Ping timeout: 240 seconds]
orivej has joined #systemtap
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
derek0883 has joined #systemtap
derek0883 has quit [Remote host closed the connection]
<agentzh> fche: well, we indeed let machines generate 99.99% of our stap scripts nowadays. manual stap scripting is just for testing and hacking on stap itself ;)
<agentzh> it's not using neural networks yet, but maybe oneday we'll get there.
<agentzh> currently the input is sql, the output is stap scripts...
<agentzh> sometimes the input is C code or lua code. oh well.
<agentzh> i meant to create some videos and blog posts to demonstrate all these soon. unforunately your stap mailing list hate youtube links. will find other ways...
<agentzh> *hates
hpt has joined #systemtap
<fche> the list takes youtube links fine
<fche> just don't dress 'em in html
<agentzh> ah, okay
derek0883 has joined #systemtap